Custom Modules

---

Custom modules can be defined either inside the KELE codebase or by importing and registering strategies from external modules. The external registration approach is recommended for most users because it avoids modifying the library itself.

I. Rule Selection Module

Option A: external module (recommended)

1. Create a Python module in your own project (for example, my_app/strategies.py).
2. Inherit from the RuleSelectionStrategy protocol.
3. Register your strategy class with the top-level registry:

from kele import register
from kele.control.grounding_selector.rule_strategies import RuleSelectionStrategy


@register.rule_selector("my_rule_strategy")
class MyRuleStrategy(RuleSelectionStrategy):
    ...

4. Reference "my_rule_strategy" via the grounding_rule_strategy configuration.

Option B: internal module

1. Create a .py file named _<name>_strategy.py, and place it under the kele/control/grounding_selector/_rule_strageties directory;
2. Inherit from the RuleSelectionStrategy class, and declare at least the functions required by this Protocol;
3. Register your strategy class with @register_strategy('<name>'). Afterwards, you can use the strategy via grounding_rule_strategy;
4. Remember to adjust the type annotation of grounding_rule_strategy (add a candidate value to the Literal).

II. Fact Selection Module

Since the engine performs reasoning at the item level, this is effectively term selection. Follow the same process as the rule selection module, but inherit from the TermSelectionStrategy protocol and register with register.term_selector(...) when defining external modules. Use the implementation via the grounding_term_strategy parameter. If you add internal strategies, remember to update the type annotation of grounding_term_strategy as well (add the new Literal candidate).